README.md

tmscoring

Python implementation of the TMscore program to compare structures of the same protein.

Usage:

We provide three classes, TMscoring, Sscoring, and RMSDscoring, that only differ in their default optimisation score.

They are initialised with the file paths to two PDB files:

alignment = tmscoring.TMscoring('structure1.pdb', 'structure2.pdb')

# Find the optimal alignment
alignment.optimise()

# Get the TM score:
alignment.tmscore(**alignment.get_current_values())

# Get the TM local scores:
alignment.tmscore_samples(**alignment.get_current_values())

# RMSD of the protein aligned according to TM score
alignment.rmsd(**alignment.get_current_values())

# Returns the transformation matrix between both structures:
alignment.get_matrix(**alignment.get_current_values())

# Save the aligned files:
alignment.write(outputfile='aligned.pdb', appended=True)

The structures can be matched by index (default), or performing a global sequence alignment with Smith-Waterman using a match score of 2, mismatch of -1, a gap penalty of -0.5 for opening and -0.1 for extending.

Utility functions:

get_tm(path_to_pdb1, path_to_pdb2) and get_rmsd(pdb1, pdb2) are simple wrappers that compute TM score or RMSD.

What is different?

tmscoring is a Python library that conveniently exposes all the necessary variables. This removes the necessity to parse files.

Also, the minimisation engine is MINUIT's Migrad, a powerful and robust derivative-free minimisation algorithm, heavily tested by particle physicists for decades. In our testing, tmscoring yields the same or slightly better scores than upstream TMscore.